A novel nomogram-based scoring system for early prediction of direct oral anticoagulant–associated bleeding risk in geriatric Chinese patients with nonvalvular atrial fibrillation–related ischemic stroke: development and temporal validation in a rivaroxaban-predominant single-center cohort

This study aimed to construct and validate a user-friendly nomogram-based scoring model for predicting the risk of all-cause bleeding in geriatric patients with non-valvular atrial fibrillation-related ischemic stroke (NVAF-IS) treated with direct oral anticoagulants (DOACs) in a rivaroxaban-predominant Chinese single-center. This real-world cohort study was conducted on 1943 consecutive hospitalized geriatric patients with NVAF-IS who received DOACs (predominantly rivaroxaban) therapy. Risk factors were identified using four feature-selection methods: lasso regression, random forest, Boruta algorithm, and backward stepwise regression. A nomogram-based scoring model was developed and evaluated for discrimination, calibration, and clinical utility. In this rivaroxaban-predominant single-center cohort (rivaroxaban 88.89%, edoxaban 6.55%, dabigatran 4.56%), a total of 1251 participants were ultimately included, of whom 920 were used for model development and 331 for temporal validation. Seven risk factors were identified: uncontrolled hypertension, diabetes mellitus, prior bleeding, anemia, concomitant antiplatelet drugs use, age, and albumin. A nomogram-based UDPA 4 score and a software-based risk calculator were constructed specifically for geriatric Chinese patients with NVAF-IS receiving DOACs (predominantly rivaroxaban), and demonstrated good discrimination (derivation AUC = 0.824; validation AUC = 0.768), calibration, and clinical utility. Our model outperformed the widely used ORBIT (AUC = 0.632), ATRIA (AUC = 0.663), and HAS-BLED (AUC = 0.630) scores in this specific population. Subgroup analyses were conducted in adults ≥ 75 years old, patients without prior bleeding, and patients receiving rivaroxaban. Sensitivity analysis was performed with age and albumin as continuous variables. All analyses achieved acceptable model performance. This study developed and validated, for the first time, a clinical risk score and a software-based risk calculator specifically for elderly Chinese patients with NVAF-IS receiving DOACs in a rivaroxaban-predominant single-center cohort. Our findings may facilitate individualized bleeding risk assessment in this specific population.